Category: New York

Quick Note: How to Incentivize Transit-Oriented Development

The Biden administration recently put out a statement saying that it would work to increase national housing production. It talks about the need to close the housing shortfall, estimated at 1.5 million dwellings, and proposes to use the Bipartisan Infrastructure Law (BIL) to dole out transport funding based on housing production. This is a welcome development, and I’d like to offer some guidelines for how this can be done most effectively.

Incentives mean mistrust

You do not need to give incentives to trustworthy people. The notion of incentives already assumes that the people who are so governed would behave poorly by themselves, and that the governing body, in this case the federal government, surveils them loosely so as to judge them by visible metrics set in advance. Once this fundamental fact is accepted – the use of BIL funding to encourage housing production implies mistrust of all local government to build housing – every other detail should be set up in support of it.

Demand conflict with community

Federal funding should, in all cases, require state and local governments to discipline community groups that fight housing and extract surplus from infrastructure. Regions that cannot or do not do so should receive less funding; the feds should communicate this in advance, stating both the principle and the rules by which it will be judged. For example, a history of surrender to local NIMBYs to avoid lawsuits, or else an unwillingness to fight said lawsuits, should make a region less favored for funds, since it’s showing that they will be wasted. In contrast, a history of steamrolling community should be rewarded, showing that the government is in control and prioritizes explicit promises to the feds and the voters over implicit promises to the local notables who form the base of NIMBYism.

Spend money in growth regions

In cities without much housing demand, like Detroit and Cleveland, the problem of housing affordability is one of poverty; infrastructure spending wouldn’t fix anything. This means that the housing grant should prioritize places with growth demand, where current prices greatly exceed construction costs. These include constrained expensive cities like New York and San Francisco, but increasingly also other wealthy cities like Denver and Nashville, whose economic booms translate to population increase as well as income growth, but unfortunately housing growth lags demand. Even poorer interior cities are seeing rent increases as people flee the high prices of richer places, and encouraging housing growth in their centers is welcome (but not in their suburbs, where housing is abundant and not as desirable).

Look at residential, not commercial development

In the United States, YIMBY groups have focused exclusively on residential development. This is partly for political reasons: it’s easier to portray housing as more moral, benefiting residents who need affordable housing even if the building in question is market-rate, than to portray an office building as needing political support. In some cases it’s due to perceived economic reasons – the two cities driving the American YIMBY discourse, New York and San Francisco, have unusually low levels of job sprawl for the United States, and in both cities YIMBY groups are based near city center, where jobs look especially plentiful. At the local and state level, this indifference to commercial YIMBY is bad, because it’s necessary to build taller in city center and commercialize near-center neighborhoods like the West Village to fight off job sprawl.

However, at the federal level, a focus on residential development is good. This is a consequence of the inherent mistrust assumed in the incentive system. While economically, American cities need city centers to grow beyond the few downtown blocks they currently occupy, politically it’s too easy for local actors to bundle a city center expansion with an outrageously expensive urban renewal infrastructure plan. In New York, this is Penn Station redevelopment, including some office towers in the area that are pretty useful and yet have no reason to be attached to the ill-advised Penn Station South project digging up an entire block to build new tracks. Residential development is done at smaller scale and is harder to bundle with such unnecessary signature projects; the sort of projects that are bundled with it are extensions of urban rail to new neighborhoods to be redeveloped, and those are easier to judge on the usual transport metrics.

Consolidating Stops with Irregular Spacing

There was an interesting discussion on Twitter a few hours ago about stop consolidation on the subway in New York. Hayden Clarkin, the founder of TransitCon, brings up the example of 21st Street on the G in Long Island City. The stop is lightly-used and very close to Court Square, which ordinarily makes it a good candidate for removal, a practice that has been done a handful of times in the city’s past. However, the spacing is irregular and in context this makes the stop’s removal a lower-value proposition; in all likelihood there should not be any change and trains should keep calling at the station as they do today.

What is 21st Street?

The G train, connecting Downtown Brooklyn with Long Island City directly, makes two stops in Queens today: Court Square, at the southern end of the Long Island City business district, and 21st Street, which lies farther south. Here is a map of the area:

Source: NYCT neighborhood map

At closest approach, the platforms of 21st are 300 meters away from those of Court Square on the G; taking train length into account, this is around 400 meters (the G runs short trains occupying only half the platform). Moreover, Court Square is a more in-demand area than 21st Street: Long Island City by now near-ties Downtown Brooklyn as the largest job center in the region outside Manhattan, and employment clusters around Queens Plaza, which used to be one stop farther north on the G before the G was curtailed to Court Square in order to make more room for Manhattan-bound trains at Queens Plaza. Court Square is still close to jobs, but 21st Street is 400 meters farther away from them, with little on its side of the neighborhood.

Stop spacing optimization

Subways cannot continuously optimize their stop spacing the way buses can. Building a new bus stop costs a few thousand dollars, or a few ten thousand if you’re profligate. Building a new subway stop costs tens of millions, or a few hundred million if you’re profligate. This means that the question of subway stop optimization can only truly be dealt with during the original construction of a line. Subsequently, it may be prudent to build a new stop but only at great expense and usually only in special circumstances (for example, in the 1950s New York built an infill express station on the 4 and 5 trains at 59th, previously a local-only station, to transfer with the N, R, and W). But deleting a stop is free; New York has done it a few times, such as at 18th Street on the 6 trains or 91st on the 1. Is it advisable in the case of 21st?

The answer has to start with the formula for stop spacing. Here is my earliest post about it, in the context of bus stops. The formula is,

\mbox{Optimum spacing} = \sqrt{4\cdot\frac{\mbox{walk speed}}{\mbox{walk penalty}}\cdot\mbox{stop penalty}\cdot\mbox{average trip distance}}

The factor of 4 in the formula depends on circumstances. If travel is purely isotropic along the line, then the optimum is at its minimum and the factor is 2. The less isotropic travel is, the higher the factor; the number 4 is when origins are purely isotropic, which reflects residential density in this part of New York, but destinations are purely anisotropic and can all be guaranteed to be at distinguished nodes, like business centers and transfer points. Because 21st Street is a residential area and Court Square is a commercial area and a transfer point, the factor of 4 is justified here.

Walk speed is around 1.33 m/s, the walk penalty is typically 2, the stop penalty on the subway is around 45 seconds, and the average unlinked trip on the subway is 6.21 km; the formula spits out an optimum of 863 m, which means that a stop that’s 400 meters from nearby stops should definitely be removed.

But there’s a snag.

The effect of irregular stop spacing

When the optimal interstation is 863 meters, the rationale for removing a stop that’s located 400 meters from adjacent stations is that the negative impact of removal is limited. Passengers at the stop to be removed have to walk 400 meters extra, and passengers halfway between the stop and either of the adjacent stops have no more walking to do because they can just walk to the other stop; the average extra walk is then 200 meters. The formula is based on minimizing overall travel time (with a walk penalty) assuming that removing a stop located x meters from adjacent stops incurs an extra walk of x/2 meters on average near the station. Moreover, only half of the population lives near deleted stops, so the average of x/2 meters is only across half the line.

However, this works only when stop spacing is regular. If the stop to be removed is 400 meters from an adjacent stop, but much farther from the adjacent stop on the other side, then the formula stops applying. In the case of 21st Street, the next stop to the south, Greenpoint Avenue, is 1.8 km away in Brooklyn, across an unwalkable bridge. Removing this stop does not increase the average walk by 200 meters but by almost 400, because anywhere from 21st south in Long Island City the extra walk is 400. Moreover, because this is the entire southern rim of Long Island City, this is more than just half the line in this area.

In the irregular case, we need to halve the factor in the formula, in this case from 4 to 2 (or from 2 to 1 if travel is isotropic). Then the optimum falls to 610; this already takes into account that 21st Street is a weaker-demand area than Court Square, or else the factor in the formula would drop by another factor of 2. At 610 meters, the impact of removing a stop 400 meters from an adjacent stop is not clearly positive. In the long run, it is likely counterproductive, since Long Island City is a growth area and demand is likely to grow in the future.

Does this generalize?


In New York, this situation occurs at borough boundaries, and also at the state boundary if more service runs between the city and New Jersey. For example, in retrospect, it would have been better for the east-west subway lines in Manhattan to make a stop at 1st or 2nd Avenue, only 300-500 meters from the typical easternmost stop of Lexington. The L train does this, and if anything does not go far enough – there’s demand for opening a new entrance to the 1st Avenue stop (which is one of the busiest on the line) at Avenue A, and some demand for a likely-infeasible infill stop at Avenue C. These are all high-density areas, but they’re residential – most people from Queens are not going to 2nd Avenue but to Lex and points west, and yet, 2nd would shorten the walk for a large group of residential riders by around 400 meters, justifying its retrospective inclusion.

No Federal Aid to Transit Operations, Please

This is the third in a series of four posts about the poor state of political transit advocacy in the United States, following posts about the Green Line Extension in metro Boston and free public transport proposals, to be followed by an Urban Institute report by Yonah Freemark.

In the United States, political transit activists in the last few years have set their eyes on direct federal aid for operating subsidies for public transport. Traditionally, this has not been allowed: federal aid goes to capital planning (including long-term maintenance), and only a small amount of money goes to operations, all in peripheral bus systems. Urban transit agencies had to operate out of fares and local and state money. Demands for federal aid grew during corona, where emergency aid to operations led to demands for permanent subsidies, and have accelerated more recently as corona recovery has flagged (New York’s subway ridership is only around 60% of pre-corona levels). But said demands remain a bad idea in the short and long terms.

In the early 20th century, when public transport was expected to support itself out of fares, operating costs grew with wages, but were tempered by improvements in efficiency. New York City Transit opened with ticket-takers at every subway entrances and a conductor for every two cars; within a generation this system was replaced with automatic turnstiles and one conductor per train. Kyle Kirschling’s thesis has good data on this, finding that by the 1930s, the system grew to about 16,000 annual car-miles (=26,000 car-km) per employee.

And then it has stagnated. Further increases in labor efficiency have not happened. Most American systems have eliminated conductors, often through a multi-decade process of attrition rather than letting redundant workers go, but New York retains them. The network today actually has somewhat less service per employee than in the 1930s, 14,000 car-miles as of 2010, because fixed costs are spread across a slightly smaller system. Compare this with JICA’s report for Mumbai Metro comparing Japanese cities: Tokyo Metro has 283,871,000 car-km (PDF-p. 254) on 8,474 employees (PDF-p. 9), which is 33,500/employee, and that’s without any automation and with only partially conductor-less operations; Yokohama gets 40,000.

Moreover, the timeline in the US matches the onset of subsidies, to some extent: state and local subsidies relieved efficiency pressure. In Canada, TTC saw this and lobbied against subsidies for its own operations in the 1960s, on the grounds that without a breakeven mandate, the unions would capture all surplus; it took until the 1970s for it to finally receive any operating subsidies.

Federal subsidies make all of this worse. They are other people’s money (OPM), so local agencies are likely to maximize them at the expense of good service; this is already what they do with capital money, lading projects with local demands for betterments figuring that if everyone else hogs the trough then they should as well.

Then there is the issue of wages. Seniority systems in American unionized labor create labor shortages even when pay is high, because of how they interact with scheduling and tiered wage structures. Bus drivers in Boston earn around $80,000 a year, a pay that German bus and train drivers can only dream of, but starting drivers are in probational status and have a lower wage (they are not even given full-time work until they put in a long period of part-time work). Moreover, because drivers pick their shifts in seniority order, drivers for about the first 10 years are stuck with the worst shifts: split shifts, graveyard shifts at inconsistent intervals, different garages to report to. New York manages to find enough bus drivers to fill its ranks but only by paying around $85,000 a year; other American cities, paying somewhat less, are seeing thousands of missed runs over the year because they can’t find drivers.

And outside aid does nothing to fix that. Quite to the contrary, it helps paper over these problems and perpetuates the labor gerontocracy. New York City Transit has learned to react to every crisis by demanding a new source of income; there is not enough political appetite for transparent taxation, so the city and state find ever more opaque sources of funds, avoiding political controversy over wanton inefficiency but creating more distortion than a broad income tax would.

Instead of subsidizing current consumption, a developmental state should subsidize production. Don’t pay money to hire more bus drivers; pay for automating subway systems, for better dispatching, for better planning around intermodal integration. Current American wages, not to mention the unemployment rate, scream “invest in labor-saving technology” and not “expand labor-intensive production.”

The G Train

The G train is bad. I say this, 16 years after I moved to New York, 11 years after I left, and I know it’s what every New Yorker knows. Tourists walk too slowly, rent is too high for small apartments, and the G train sucks. What I want to highlight in this post is how the subway’s scheduling paradigm is especially bad for the G train and leads to a vicious cycle making the train less frequent and less useful for passengers.

The role of the G train

The G train is the only mainline subway service in New York that does not enter Manhattan; see map here. It connects what are now the region’s two largest non-Manhattan business centers, Long Island City and Downtown Brooklyn, running vaguely parallel to the East River on the Queens and Brooklyn side of it. To the south of Downtown Brooklyn, it has a tail serving the wealthy neighborhoods collectively called South Brooklyn, such as Carroll Gardens and Park Slope.

I’ve criticized the G before for its poor construction. It misses critical transfers, like the other lines built in the IND program in the 1920s-30s. In Queens it misses Queensboro Plaza and the transfer to the N/W trains on the Astoria Line, and in Brooklyn it misses every single non-IND line except the L (and, at a suboptimal location, the R). This already makes it less useful as a circumferential line – such lines live on convenient transfers to radial lines, because direct O&D service is less valuable to secondary destinations than to primary ones.

But what I realized last week, commuting from Long Island City to Downtown Brooklyn, is more delicate. My hotel was near Queensboro Plaza, which the G doesn’t serve, but the station is served by the 7, which connects to the G one stop away at Court Square; Marron’s new office is in Downtown Brooklyn right on top of the Jay Street station, on the IND-built A/C and F trains, which is either a cross-platform connection or a short walk from the G. So for my trip, the connections worked. And yet, I was regularly facing 10-minute waits on the shoulders of rush hour, and on the subway countdown clock I saw a 15-minute gap.

To explain what went so wrong that the G should have such low frequency at 10 in the morning, it’s necessary to explain how New York City Transit decides the frequency of each service during each time of day.

New York City Subway frequency

In New York, the system for deciding the frequency of each subway service at each time of day is based on average peak crowding. This means that for all trains using the service in a given time period, the crowding level at the peak crowding point of the journey is averaged; frequency is adjusted so that off-peak the peak crowding level is 125% of seated capacity, and at rush hour it is based on published standing capacity per car that works out to about 300% of seated capacity depending on car design.

This system is done per numbered or lettered service. Thus, for example, the 2 and 3 trains run on the same track most of the way, but where they diverge, the 2 is considerably busier, and therefore the 2 runs slightly higher frequency (most ridership on the 2 and 3 is on the shared segment, not the tails). As a result, on the shared trunk, there cannot be perfect alternation of 2 and 3 trains; a few times an hour, a 2 train is followed by another 2 train, which means that on the tail, the frequency is uneven. When two 2 trains follow each other with no 3 between them, the leading 2 train is more crowded than the trailing one; this variation is averaged out in the guidelines – it is not the busiest train that sets the frequency guidelines.

These guidelines are not a good way to timetable trains. The above example of how it can create uneven crowding on the 2 is one problem with this system; if instead there were regular alternation of 2 and 3 trains then the 2 would be persistently slightly more crowded than the 3, just as today there is uneven crowding whenever two 2 trains run with no 3 in between, but the frequency on both the shared trunk and the branches would be more regular. This is especially important on more complexly interlined parts of the network, where the current system leads to large programmed gaps between trains occasionally.

The G is not very heavily interlined; the issue there relates to another criticism of the guidelines, which is that they assume travel demand is fixed. If the ridership on a train is independent of frequency, which it is if the headway between trains is very short compared to the trip time (say, if the trains run every 2-3 minutes), then the sole purpose of service is to provide the capacity the passengers need, and so the guidelines make sense as a way of rationing service convenience. However, in reality, the elasticity of ridership with respect to service provision is not zero. Three years ago I did some analysis of New York’s situation and the existing literature on ridership-frequency elasticity, suggesting it is equal to about 0.4. So the low frequency of the G deters ridership, which then appears to justify the low frequency.

But 0.4 < 1. And I believe that there are two reasons why on the G, and on circumferential lines in general, the elasticity of ridership with respect to frequency should be higher.

Trip length

Circumferential lines in general tend to have shorter average trip time. Between two nearby spokes, say between Downtown Brooklyn and Williamsburg, they are the only real option; between two farther away ones, a direct radial may be an alternative.

The G is different from (say) the Ringbahn in that it misses most transfers, but this should not impact this pattern too much. The missed transfers in Downtown Brooklyn weaken the G for short as well as long trips involving a connection there. In contrast, in the middle the G does make the most important transfer, that with the L, and only misses the weaker J/M/Z.

The 0.4 estimate for ridership elasticity with respect to frequency assumes average behavior for trip length. But if trips are shorter, then the impact of frequency is larger. The 0.4 estimate comes out of an estimate of about -0.8 of ridership with respect to generalized trip time, which includes in-vehicle time, walk time, and wait time, the latter two given extra weight to account for transfer penalty. If one of the three components of trip times is shortened, the other two grow in importance.

The role of options

The G is not usually passengers’ only choice for making the trip. They can connect in Manhattan, or, in some cases, go directly via Manhattan, for example taking the N or R from Downtown Brooklyn to Queens (in the opposite direction, they serve separate station so it’s a harder choice, leading to asymmetric demand). Going between Marron and the East Village, Eric Goldwyn could connect to the L via the A/C/F or the G; I never once saw him use the G, only the lines via Manhattan.

I have not seen the impact of different transit paths on demand elasticity in the literature. It is likely that the elasticity in such case must be higher, because it is standard in economics that demand is more elastic for goods sold on a competitive market than by a monopolist.

Note also that it is to the overall system’s benefit to convince passengers to switch from radial lines to the G. The G is less crowded, so such a switch distributes ridership better on the system. And the G starts out much less frequent, so that even on a fixed operating budget, the impact of a service increase on the G on ridership is larger than on an already frequent trunk.

How High-Speed and Regional Rail are Intertwined

The Transit Costs Project will wrap up soon with the report on construction cost differences, and we’re already looking at a report on high-speed rail. This post should be read as some early scoping on how this can be designed for the Northeast Corridor. In particular, integration of planning with regional rail is obligatory due to the extensive track sharing at both ends of the corridor as well as in the middle. This means that the project has to include some vision of what regional rail should look like in Boston, New York, Philadelphia, and Washington. This vision is not a full crayon, but should have different options for different likely investment levels and how they fit into an intercity vision, within the existing budget, which is tens of billions thanks to the Bipartisan Infrastructure Framework.


In Boston, commuter rail and intercity rail interact via the Providence Line, which is double-track. The Providence Line shares the same trunk line into Boston with the Franklin Line and the Stoughton Line, and eventually with South Coast Rail services.

The good news is that the MBTA is seriously looking at electrifying the trains to a substantial if insufficient extent. The Providence Line is already wired, except for a few siding and yard tracks, and the MBTA is currently planning to complete electrification and purchase EMUs on the main line, and possibly also on the Stoughton Line; South Coast Rail is required to be electrified when it is connected to this system anyway, for environmental reasons. If there is no further electrification, then it signals severe incompetence in Massachusetts but is still workable to a large extent.

Options for scheduling depend on how much further the state invests. The timetables I’ve written in the past (for an aggressive example, see here) assume electrification of everything that needs to be electrified but no North-South Rail Link tunnel. An NSRL timetable requires planning high-speed rail in conjunction with the entirety of the regional rail system; this is true even though intercity trains should terminate on the surface and not use the NSRL tunnel.


Philadelphia is the easiest case. Trenton-Philadelphia is four-track, and has sufficiently little commuter traffic that the commuter trains can be put on the local tracks permanently. In the presence of high-speed rail, there is no need for express commuter trains – passengers can buy standing tickets on Trenton-Philadelphia, and those are not going to create a capacity crunch because train volumes need to be sized for the larger peak market into New York anyway.

On the Wilmington side, the outer end of the line is only triple-track. But it’s a short segment, largely peripheral to the network – the line is four-track from Philadelphia almost all the way to Wilmington, and beyond Wilmington ridership is very low. Moreover, Wilmington itself is so slow that it may be valuable to bypass it roughly along I-95 anyway.

The railway junctions are a more serious interface. Zoo Interlocking controls everything heading into Philadelphia from points north, and needs some facelifts (mainly, more modern turnouts) speeding up trains of all classes. Thankfully, there is no regional-intercity rail conflict here.


In some ways, the Washington-Baltimore Penn Line is a lot like the Boston-Providence line. It connects two historic city centers, but one is much larger than the other and so commuter demand is asymmetric. It has a tail behind the secondary city with very low ridership. It runs diesel under catenary, thanks to MARC’s recent choice to deelectrify service (it used to run electric locomotives).

But the Penn Line has significant sections of triple- and quad-track, courtesy of a bad investment plan that adds tracks without any schedule coordination. The quad-track segment can be used to simplify the interface; the triple-track segment, consisting of most of the line’s length, is unfortunately not useful for a symmetric timetable and requires some strategic quad-track overtakes. The Penn Line must be reelectrified, with high-performance EMUs minimizing the speed difference between regional and intercity trains. There are only five stations on the double- and triple-track narrows – BWI, Odenton, Bowie State, Seabrook, New Carrollton – and even figuring differences in average speed, this looks like a trip time difference between 160 km/h regional rail and 360 km/h HSR of about 15 minutes, which is doable with a single overtake.

New York

New York is the real pain point. Unlike in Boston and Washington, it’s difficult to isolate different parts of the commuter rail network from one another. Boston can more or less treat the Worcester, Providence+Stoughton, Fairmount, and Old Colony Lines as four different, non-interacting systems, and then slot Franklin into either Providence or Fairmount, whichever it prefers. New York can, with current and under-construction infrastructure, plausibly separate out some LIRR lines, but this is the part of the system with the least interaction with intercity rail.

Gateway could make things easier, but it would require consciously treating it as total separation between the Northeast Corridor and Morris and Essex systems, which would be a big mismatch in demand. (NEC demand is around twice M&E demand, but intercity trains would be sharing tracks with the NEC commuter trains, not the M&E ones; improving urban commuter rail service reduces this mismatch by loading the trains more within Newark but does not eliminate it.)

It’s so intertwined that the schedules have to be done de novo on both systems – intercity and regional – combined. This isn’t as in Boston and Washington, where the entire timetable can be done to fit one or two overtakes. This isn’t impossible – there are big gains to be had from train speedups all over and there. But it requires cutting-edge systems for timetabling and a lot of infrastructure investment, often in places that were left for later on official plans.

Quick Note: the LaGuardia Transit Connector

It’s amazing how much good can happen when an obstacle like Andrew Cuomo is removed. In lieu of his backward air train proposal, hated by just about everyone not on his payroll, Governor Kathy Hochul is moving forward on a better set of alternatives for a mass transit connection to LaGuardia. It’s interesting to see what the process is looking at but also what it isn’t; so far this looks better than the alternatives analysis for Interborough Express (ex-Triboro).

So far I have not seen analysis, only drawings of 14 alternatives. As with the IBX study, the LGA plan distinguishes different modes of public transit – there are bus, light rail, subway, and even ferry options. But it doesn’t stop there. It looks at multiple alignments: the scope is how to connect LGA to the rest of the city the best, and this can be done from a number of different directions – even a backward train (as light rail) along an alignment similar to Cuomo’s is present, and will likely not advance further because of its circuitous route.

Among the 14 alternatives, I think the obviously best one is a subway extension (slide 12 above); another subway option, a branch following the Grand Central Parkway (slide 11), is inferior because of branching splits frequencies and ridership at the cut off Astoria-Ditmars Boulevard station is high. A subway extension promises a connection in around 30 minutes to Times Square, every 5 minutes all day, with good connections to other destinations via the transfers at Queensboro Plaza and in Midtown.

The one thing that I’m sad the analysis hasn’t looked at is intermediate stations. It’s around 4.5 km from Ditmars to the main LGA terminal along the proposed alignment, passing through redevelopable industrial land and through residential land in Astoria Heights awkwardly tucked between airport grounds and Astoria proper. The same quality of service that the airport could get, these neighborhoods could get as well, except a hair faster because they’re closer.

Extending the Astoria Line is especially useful since it is short and not especially crowded until it hits Queensboro Plaza and inherits the crowding of the 7 train and its riders. In the context of deinterlining the subway, this is especially valuable: right now 60th Street Tunnel carries the N and W from Astoria but also the R from Queens Boulevard, and under deinterlining the tunnel would carry only Astoria riders, and so to match the high demand to 60th Street it’s valuable to create as much ridership as possible on the Astoria Line past Queensboro Plaza.

I hope that the alternatives analysis considers multiple stopping patterns in the future – that is, not just a nonstop route from Ditmars to the airport, but also an option with intermediate stations. (This does not mean local and express trains – either all trains should run locals, or all should run nonstop.) The cost of those stations is not high as it’s an elevated line, and the stop penalty on the subway is less than a minute since the top speed is so low (it looks like 45 seconds in practice comparing local and express trains on the same line).

Penn Station Tracks

In 2015, I argued that New York Penn Station should be replaced with a hole in the ground, and such a station would have sufficient capacity. I will defend those posts: in the 21st century, elaborate stations are not required for high-quality rail service, and it’s more important to have good passenger egress and intermodal connections than a signature station. The topic of this post is more niche: which rail lines should connect to Penn Station?

The three-line system

In all writing I’ve done on the subject since around 2010, I’ve assumed that Penn Station should be a three-line stations. In blog posts about regional rail for New York I’ve consistently called them Lines 1, 2, and 3; one map can be found in this post, with slightly less expansive version on Google Maps, and, consistently, Line 1 (red) is the existing Northeast Corridor, Line 2 (green) runs along the same route but uses the Gateway tunnel across the Hudson and then goes via Grand Central, and Line 3 (orange) connects the Empire Connection to the LIRR via a slightly realigned approach, otherwise using existing tracks.

At the station, their order from south to north is 2, 1, 3; the numbers are chronological (1 preexists, 2 is a higher priority to build than 3). Gateway is to enter Penn Station south of the existing tunnel and the room for a Grand Central connection is to the south (31st Street), forcing that line to be the southernmost. The East River Tunnels go under 32nd and 33rd, each as a track pair going in opposite directions rather than 32nd running eastbound and 33rd westbound, and the track pair under 33rd has a better connection to the LIRR while that under 32nd has a better connection across the Hudson; the Empire Connection loops under the Hudson tunnel to connect to southern tracks, but that’s a single-track link and needs to be doubled anyway, so it might as well be realigned.

With three lines and six approach tracks, Penn Station should have 12 platform tracks: each approach track should split into two and the two tracks should serve the same platform, a solution used for the expensive but operationally sound Stuttgart 21 project. There should not be any flexibility, save perhaps some emergency crossovers at the station, not to be used in service: the required throughput is so extensive that such flexibility is fake, reducing capacity by almost as much as the full closure of a track.

The footprint of the station looks around 155 meters wide gross, or around 145 net, corresponding to 24 per platform. The total width of the tracks is 1.7 (track center to platform) plus 4.5 (distance between track centers; Shinkansen regulations say 4.3) plus around 2 if a safety zone between each track pair is desired, which is a total of about 8 meters. The platform width is then 24 – 8 = 16. If a heavy column between two tracks adjacent to different platforms is required, this adds about another meter to maintain the safety zones, for a total of 9, resulting in 15-meter platforms.

15-meter platforms are extremely wide. Châtelet-Les Halles’s RER A and B platforms are 17 meters, and are wider than necessary; they in contrast have insufficient vertical circulation at rush hour. At 15 meters, there’s room for six escalators per access point and possibly also a staircase; at 16, there’s definitely room for the staircase. Six escalators can run without any rush hour variation, always three up and three down, and would still clear a full train with many standees in a minute. I do not foresee any capacity problems at the station if it is built this way.

But this leads to the question: since the platforms are so oversize, perhaps it is useful to have more of them at lower width?

The four-line system

Penn Station could potentially serve not three lines but four. Right now it only has infrastructure for a line and a half, and with Gateway it would have one and two halves; even three looks like a generational project. But there’s good cause to think even farther ahead and make room for a fourth line: a dedicated intercity railway. The four-line system would maintain Lines 1, 2, and 3 as above, but then add an unnumbered line with no regional trains, only intercity train.

This comes out of my ridership model for high-speed rail for the United States: at full buildout, the system would be difficult to fit into an approach track with regional trains, and regional trains would only be able to run every 5 minutes or even worse, rather than every 2 or 2.5. Moreover, once high-speed rail exists on the Northeast Corridor, the return on investment on extensions is so great that it is likely that such extensions will happen. Politics make such extensions even more favorable: high-profile investment in the Northeast’s intercity rail and in New York is likely to lead to demand for such investment in other regions, regardless of the business case, and it is fortunate that the business case for such extensions is strong independently of the politics.

I presume that, from south to north, the platform order should be Line 2 eastbound, Line 2 westbound, intercity eastbound, Line 1 eastbound, Line 1 westbound, intercity westbound, Line 3 eastbound, Line 3 westbound. The problem here is that Penn Station’s footprint is only adjacent to three east-west streets, not four, and so the intercity tunnels have to duck under private property, and the best place for them going east is to act as 31.5th and 32.5th Streets. Using the existing tunnels and then displacing regional rail to new tunnels is also possible, but less desirable: the existing tunnels have small diameter, and so it’s easier to keep them lower-speed while the new tunnels get to be bigger and support 200 km/h while maintaining enough free air to avoid creating pressure problems in passengers’ ears.

Under this system, the existing footprint of Penn Station is wide enough for 18 meters gross per each of the eight platforms, or 10 meters net. This is not out of the question, and would ordinarily be completely fine: it’s enough for four escalators per access point, or three and a staircase. At Penn Station I am slightly squeamish purely because on Lines 1 and 3 it’s the only city center station, and thus more crowded than the usual for a regional train station.

But it’s possible to slightly widen the footprint. Under no circumstances should there be any digging past the footprint of 31st and 33rd Streets: the cost of construction under existing buildings is too high. Plans for demolishing the block between 30th and 31st Streets (Block 780) are in an advanced stage, related to both a real estate deal with Vornado and plans for Penn Station South expansion, but they are extraordinarily expensive (around $10 billion at this point), and redevelopment of the block is easier on firma than over rail tracks. For all intents and purposes, the maximum usable footprint is between the lot lines of 31st and 33rd, which is 175 meters gross, perhaps 160 net with some distance between the dig and the lot line.

With 160 net meters, there are 20 meters per platform with tracks, or 12 per platform alone. This is wide enough for anything: four escalators and a staircase fit, which has enough capacity (albeit with some compromises) with permanent escalator directionality and more than enough if escalators run three-and-one at rush hour.

The benefits of creating about two extra meters per platform should be weighed against the cost of adding to the footprint of Penn Station, which is not $10 billion but also not zero, and I don’t want to make pronouncements without seeing a reliable estimate. This also depends on the difficulty of building intercity rail tunnels under private property.

Coordinated planning

A coordinated Penn Station rebuild plan should be considered together with some plan for how to use those tracks. Infrastructure investment must always come with a precise service plan, with sample timetables to the minute shared with the public for democratic review.

The upshot is that Penn Station rebuild must come with a good idea of how much service the region expects to run. A high-speed rail plan argues in favor of the four-line system, provided the cost of the extra tunnels is reasonable (low-to-mid single-digit billions; $10 billion is far too high). Otherwise, the three-line system is better.

Radial Metro Design on Rivers

The most common and most useful design paradigm for an urban metro system is radial. Subway lines should be running across the city, passing through city center with transfers to other radial lines; larger cities can also support a circumferential line, or for the largest megacities (like Moscow) two, and unless there are multiple circumferentials, every pair of lines should intersect with a transfer. For example, here is Prague:

There are three lines, meeting in a Soviet triangle, running from one side of city center to the other. Together with an intact tramway network, this boosts Prague’s annual urban rail ridership to around 830 million a year, which is 310/capita, a figure that isn’t far lower than Tokyo’s and is higher than anywhere else I can think of.

But in some cities – but not Prague – there’s a kink in the radial design. For example, here’s Kyiv, with planned expansion:

The three existing lines form a perfect Soviet triangle. Line 4, Podilsko-Vyhurivska, is under construction and radial as well. And then there is the under-construction eastern extension of Line 3, Syretsko-Pecherska, looping back to meet Line 1 at Darnytsia. This is not standard radial design. But it’s fully understandable given the situation of Kyiv.

Kyiv has a division into left-bank and right-bank Kyiv. The Dnipro is, with islets included, 1-2 km wide, one of the widest rivers of Europe. There are few bridges. The main of the city is on the right bank, but left-bank Kyiv has its own independent center around Darnytsia, encouraged by the city’s development plan precisely because the river is such an obstacle.

The river division is not universal. Prague doesn’t quite have it – the Vltava is 160-200 m wide and there are many bridge crossings, so even though city center grew along the right bank, much of the near-center is on the left bank. The city is also hilly enough that there’s no coherent left- vs. right-bank identity, and the streetcar system is sufficient to connect left-bank neighborhoods with each other without passing through city center.

Conversely, London does have this division. Bank terms are not used there – one says North and South London – but the situation is the same, even though the Thames at 250 meters is not much wider than the Vltava, and has many crossings as well. Nonetheless, a South London identity exists, defined by paucity of river crossings to East London (but not to Central or West London), and by its own centers at Waterloo and London Bridge.

As a result, the radial Underground network forms a coherent sub-network in South London. Just as the Kyiv Metro is planned to feature a loop back on Line 3 in left-bank Kyiv starting 2023, London built the Victoria line to swerve east to cross each trunk of the Northern line twice, once in North London and once in South London, and the crossing with the main line at Stockwell is even cross-platform. Unfortunately, the South London crossing with the Battersea extension is without a transfer, a deliberate design decision made to reduce ridership and perhaps reduce crowding on the Vic.

Finally, New York should think explicitly in terms of right- and wrong-side parts of the city, the right side referring to city center, that is Manhattan. New York’s subway network is not radial, but the same principles apply just the same. There is a strong wrong-side identity for Brooklyn, and historically Downtown Brooklyn was a very large business center; today it remains near-tied with Long Island City for largest job center in the region outside Manhattan. Early-20th century designers did not think in such comparative terms but they understood that it was valuable to connect Brooklyn homes with Brooklyn jobs, and thus most subway lines in Brooklyn converge on Downtown Brooklyn, and only the J/M/Z and the L go directly from Williamsburg to Manhattan.

By a fluke, all four subway lines in Queens connect to Manhattan via Long Island City, the nearest neighborhood to Midtown. Thus, a business center emerged there, growing to rival Downtown Brooklyn; just as the city’s geography can create a subway network, the subway network can create the city’s geography.

Why is Princeton Trying to Downgrade the Dinky?

Regular users of the Northeast Corridor in New Jersey know that there is a short branch off the line serving Princeton. Mainline trains do not use it – they continue between New York and Trenton – but a two-car shuttle, affectionately called the Dinky, connects the city with the train station. Historically, this is because the Northeast Corridor in New Jersey is a then-high-speed rail cutoff from 1863, which cut off Princeton from the old line. Trains run back and forth, with timed connections between New York (but not Trenton) and Princeton.

The Princeton stop on the Dinky, as can be seen in the satellite image, lies just outside the historic municipal limits of Princeton (since merged with the surrounding township). It serves the university fairly well, but is 800 meters at closest approach to the town’s main street, Nassau Street. So there has been a study for what to do to improve city access, in which a tram-train option was studied, looked good, and was dropped anyway. There are two options left: status quo, and a downgrade of the right-of-way to light rail with buses using the same corridor.

Unfortunately, transit advocates I respect, like Sandy Johnston, think the downgrade is an upgrade. So let me explain why in fact the light rail and bus option is inferior to current commuter rail operations.

The current use of the Dinky is as a connector to the Northeast Corridor. There is approximately nothing else at Princeton Junction: it’s one of the two busiest suburban stations in New Jersey, but like the other top station, Metropark, it’s a park-and-ride, designed exclusively for car-train interface. People who ride the Dinky do so to get to New York.

This means that the timed transfer with the mainline trains is critical. Frequency on the Dinky is irrelevant: all ridership from Princeton Junction into the town is going to be on the first train or bus after the mainline trains arrive, and almost all ridership to the junction is going to be on the last train that makes the connection. While frequency is not important except insofar as it matches that of the mainline, on-train capacity is important. My 2015 recollection is that off-peak ridership on the Dinky is maybe enough to fill an articulated bus (which New Jersey Transit only runs in Newark), maybe enough for a standard bus, depending on time of day – standees are likely, and standing on a bus is an awful passenger experience. At rush hour, the Dinky runs three-car trains (update 2022-2-18: no, it’s two-car trains) and they’re full.

The timed transfer is so important that the discussion of how to improve service must center how to make the transfer more efficient. The ideal improvement should be to regularize the timetable on the mainline commuter trains, and ensure that trains in opposite directions serve Princeton Junction around the same time (this is called a knot) so that the Dinky can connect to Trenton too, and even to Philadelphia with another timed transfer at Trenton or even through-service if that fits the New Jersey Transit and SEPTA schedules.

Sandy points out to me that while the Dinky only connects Princeton with the mainline, the right-of-way of the Dinky can serve more destinations – namely, the Route 1 job cluster, visible on the map as a line of office parks.

However, bus service from town to Route 1 is unlikely to succeed. It’s going to struggle to run sufficient frequency for what it needs, even as lower-frequency rail is sufficient for the Dinky’s current role:

  • Route 1 is not on the way between town and the station – there would have to be separate buses to Route 1 from the service to the train station (which I presume will stay on rail even if the downgrade is picked). This means there’s no bundling of destinations – the buses to Route 1 have to live off of Princeton-Route 1 trips.
  • Route 1 is a freeway with destinations located somewhat away, at automobile scale. Buses can stop on the side of the road but the walk is not great on the same side of the road and hostile and unsafe if crossing the road is required. A more pleasant experience is only possible if buses turn onto side roads, splitting frequency or increasing trip times.
  • Route 1 is not a large job center. OnTheMap says that between the route of the Dinky and the junction with I-295 beyond the above satellite image, which ends at Quakerbridge Road, there are 21,000 jobs. The origins of those jobs are dispersed – only 5,000 come from within the county, and only 368 come from within Princeton.
  • Conversely, the short distance traveled means that high frequency is crucial. A one-way trip from the townhouses just north of Nassau Street to the center of the Route 1 cluster along the right-of-way of the Dinky is 5.5 km, which at BRT and freeway speed is around 10 minutes one-way; a bus running less than once every 10 minutes might as well not run – but there is no chance for such a bus to fill at current demand.

Of course, the analysis of Route 1 assumes current development patterns stay with no or moderate change. A bigger change, such as greater development along Route 1 with sprawl repair, can make this option pencil out; O&D volumes need to rise by a factor of 3 assuming 100% transit modal split, or more if modal split is lower (which it invariably is, Route 1 is not Manhattan).

But then that raises the question – why engage in development in sprawl around a plan to downgrade a rail service?

If sprawl repair is plausible, then make Princeton more bikable and then set up bike lanes on Route 1 so that people can cycle to Route 1 jobs. The same bike lanes can also connect to the Dinky, with bike parking at the station, or even potentially at Princeton Junction if it’s faster to bike those 4 km than to ride a train and transfer. In the long run, all buses are going to have to be replaced by bikes anyway – bus operating costs are only going to go up.

And if redevelopment is plausible, look again at the satellite image and see what the land use at the existing train stations is like. Princeton is one of the most expensive places in the United States, and the Dinky station has a golf course on one side; that’s 0.5 km^2 of land, or, as I prefer to think of it, 50,000 housing units. Another 0.05 km^2 consists of parking lots right near the station, and can and should be redeveloped as a town center extension for a population that can swamp the existing town population by a factor of 4. The parking lots at Princeton Junction and the undeveloped land between them are another 0.4 km^2 of prime real estate.

In general, I cannot think of any railway where service would be improved by a downgrade from mainline rail to bus. But the Dinky has specific issues making such a downgrade especially deleterious for current users, namely the need for a timed connection, while the proposed source of new trips, namely Route 1, is too weak to be worth much. Thankfully, a no-build option keeping the status quo is still under consideration, and I hope that the region chooses it and invests in making the Dinky better rather than in replacing it.

The Interborough Study

I was excited about the idea of Interborough Express (IBX) as announced by New York Governor Kathy Hochul, and then last week her office released a preliminary report about the alternatives for it, and I got less excited. But it’s not that the study is bad, or that Hochul is bad. Rather, the study is a by the numbers alternatives analysis, shorter than the usual in a good way; its shortcomings are the shortcomings of all American planning.

The main rub is that the report looks at various options for the IBX route, broken down by mode. There’s a commuter rail option, which bakes in the usual bad assumption about commuter rail operations, including heavier trains (lighter trains are legal on US tracks as of 2018) and longer dwell times that are explained as a product of the heavier trains (dwell times have nothing to do with train mass). That’s par for the course – as we saw yesterday, everything that touches mainline rail in North America becomes stupid even in an otherwise understandable report.

But even excluding commuter rail, the study classifies the options by mode, focusing on bus rapid transit and light rail (and no subway, for some reason). It compares those two options and commuter rail on various measures like expected ridership and trip times. This is normal for American alternatives analyses for new corridors like IBX: they look at different modes as the main decision point.

This is also extraordinarily bad governance. There are some fundamental questions that are treated as afterthoughts, either not studied at all or mentioned briefly as 1-2 sentences:

  • How far north should the line go? The IBX plan is to only go from Jackson Heights to the south, in contrast with older Triboro proposals going into the Bronx.
  • What should the stop spacing be? The stops can be widely spaced, as in the current proposal, which stops mainly at intersection points with other lines, or more closely spaced, like an ordinary subway line.
  • Under a light rail option, should the line be elevated where the trench is too narrow or at-grade?
  • Should freight service be retained? What are the benefits of retaining freight rail service on the Bay Ridge Branch and what are the incremental costs of keeping it versus taking over the right-of-way?
  • How large should the stations be?
  • How frequent should the trains be? If freight service is retained, what frequencies are compatible with running freight on the same tracks for part or all of the line?

A better study must focus on these questions. Some of them, moreover, must be decided early: urban planning depends on whether the line goes into the Bronx or not; and industrial planning depends on what is done with freight service along the corridor.

Those questions, moreover, are more difficult than the modal question. A BRT option on a rail corridor without closely parallel arterial roads should be dismissed with the same ease that the study dismisses options not studied, and then the question of what kind of rail service to run is much less important than the scope of the project.

But American planning is obsessed with comparing public transit by mode rather than by corridor, scope, or any other aspect. Canadian planning has the same misfeature – the studies for the Broadway SkyTrain extension looked at various BRT and light rail options throughout, even though it was clear the answer was going to be SkyTrain, and omitted more fundamental questions regarding the cost-construction disruption tradeoff or even the scope of the project (the original studies from 2012 did not look at truncating to Arbutus, an option that had been talked about before and that would eventually happen due to cost overruns).

So overall, the IBX study is bad. But it is interestingly bad. Andrew Cuomo was a despicable governor who belongs in prison for his crimes. Less criminal and yet similarly loathsome people exist in American public transit. And yet, Hochul and her office are not like that, at all. This is not a sandbag, or a corrupt deal. It’s utterly ordinary in its failure; with all the unique failures of the Cuomo era stripped, what is left is standard American practice, written more clearly than is usual, and it just isn’t up to par as an analysis.

Hochul has been moving on this project very quickly, and good transit advocates should laud this. It should not take long to publish a report comparing alternatives on more fundamental questions than mode, such as scope, the role of freight, and the extent of civil infrastructure to be used. The costs and benefits of IBX heavily depend on the decisions made on such matters; they should not be brushed aside.